Conversation
jarkko

Jarkko Sakkinen

Edited 1 year ago

I’ve opened up #development #git tree for in-kernel #Keystone #enclave #driver:

https://git.kernel.org/pub/scm/linux/kernel/git/jarkko/linux-riscv-keystone.git/

Existing branches:

  • keystone: follows Linus’ tree.
  • keystone-5.19.y: I’m actively trying to get Genesys2 FPGA hosting CVA6 working with this kernel version. Buddy allocator works but the use of dma_alloc_coherent() to allocate contiguous physical memory for enclaves freezes the system when it is being used (not in the call itself).

Keystone enclaves are applications hosted by Keystone security monitor running inside #OpenSBI, and partitioned and protected by physical memory protection (usually shortened as PMP) machine registers.

PMP is part of standard RISC-V ISA, which means that Keystone enclaves are usable with any RISC-V CPU with MMU. There is nothing that would prevent this coming eventually a mainline feature because it works for without non-standard or proprietary components involved.

Another favorable element of having Keystone enclaves as kernel feature is that Keystones enclave can run both in U-mode and S-mode meaning that kernel subsystems could use the feature too (if wanted). For instance, #Intel #SGX can host only ring-3 enclaves, which limits its scope.

#linux #kernel #riscv #cva6

2
0
2
jarkko

Jarkko Sakkinen

Reply to @jarkko
Edited 1 year ago

I can request riscv-keystone@vger.kernel.org if there is broader interest on this work.

@monsieuricon

0
0
0
conor

Conor Dooley

Reply to @jarkko
@jarkko kinda random thought of no real merit at this stage of your develop, why's the code in arch/riscv? It seems as if it would be better off in drivers/firmware?
Obv. that doesn't matter at this remove, since you're using experimental SBI extension IDs etc so nothing here is mergeable, but I am curious :)
1
0
0
jarkko

Jarkko Sakkinen

Reply to @conor
@conor it is highly CPU architecture dependent feature and any TEE must be part of the boot flow so I don't think it should linkable as module at all. As part of the boot flow any possible in-kernel use cases are enabled, i.e. other subsystems could have Keystone backend. Thus, I think it is best pick for now.
1
0
0
conor

Conor Dooley

Reply to @jarkko
@jarkko I don't think any of those things preclude being in drivers/firmware.
1
0
0
jarkko

Jarkko Sakkinen

Reply to @conor
@conor yeah, the honest truth is that I just place it somewhere :-) i think the appropriate place if anywhere in the mainline would drivers/staging for the moment.
1
0
0
conor

Conor Dooley

Reply to @jarkko
@jarkko cos of the experimental SBI extension? FWIW, the patch acceptance policy for standard extensions is that you either need to be a frozen extension from RVI or actually available to people for vendor stuff. See
<https://docs.kernel.org/riscv/patch-acceptance.html#submit-checklist-addendum>.

I don't know which of the two your keystone work is intended to be though.
1
0
0
jarkko

Jarkko Sakkinen

Reply to @conor
@conor yeah, I mean for staging afaik the main requirement is that there is someone keeping the driver up to date but otherwise it is not that conservative. so feels like a good location at this point.
1
0
0
jarkko

Jarkko Sakkinen

Reply to @jarkko
Edited 1 year ago
@conor I would like to also point out that Keystone driver is better to be a subsystem in vmlinux because it offers broader scope of applications than e.g. SGX or SEV-SNP, i.e. kernel could run payloads there and thus other subsystems could find use for it. For instance, we could possibly have backend for trusted keys and stuff like that. But for now either my korg tree or staging at most are the most appropriate...
1
0
0
jarkko

Jarkko Sakkinen

Reply to @jarkko
@conor Generally for any TEE using it as a module has zero real world applications.
1
0
0
jarkko

Jarkko Sakkinen

Reply to @jarkko
@conor ... and thus it is a non-issue for me if this cannot go to e.g. staging. The priority is to get the Keystone project use a kernel patch instead of OOT driver. While at it, it really does not hurt to ask from @gregkh if this can be picked to staging but not a thing thah I'm go to do today or tomorrow. Still because this works on standard hardware, it *does* make sense for Linux *at some point* (and open source firmware extension).

Seems much less difficult exercise than the one I had with Intel SGX in the past...
1
0
0
gregkh

Greg K-H

Reply to @jarkko
@jarkko @conor drivers/staging/ is for code that is abandoned by companies and the community needs others to clean/fix it up, it is not for "code just not ready to be merged yet" as it ends up taking more work to get code out of staging into the real part of the kernel than to just do the real work to start with out-of-tree and then submit it to the proper location.

In other words, drivers/staging/ is not a dumping ground for experiments :)

Just do it right, and get it merged upstream correctly.

And why 5.19? That's a long obsolete and insecure kernel version, no one should ever still be using that mess, especially anything purporting to have anything to do with "security" in any form.
1
0
0
jarkko

Jarkko Sakkinen

Reply to @gregkh
@gregkh @conor OK, fair enough. Project was basing on v5.19 so it was easiest to start with that. I've tried it recently with the recent kernels and it works just fine. Since it is not nearly as complicated as e.g. SGX is, I just maintain my tree up until Keystone is a stable extension.
1
0
0
jarkko

Jarkko Sakkinen

Reply to @jarkko
@gregkh @conor Stable firmware extension to be exact. It works on a standard RISC-V CPU with MMU. It even works on my VisionFive 2 RISC-V board: https://github.com/keystone-enclave/keystone/issues/339. It is more like in the area of TrustZone than big cloud things like SGX or SEV-SNP, providing tools to implement secure boot and stuff like that. Thus, I see worth of doing this on side up because I really believe that it is a piece that RISC-V needs something like this to be successful. I'll move forward in upstreaming if it becomes official part of OpenSBI firmware.
1
1
0
conor

Conor Dooley

Reply to @jarkko
@jarkko @gregkh just to note, it does not have to be part of OpenSBI. It just needs to be a frozen SBI extension. OpenSBI is but one one SBI implementation, and being part of it is not a requirement for patch acceptance.
1
0
1
jarkko

Jarkko Sakkinen

Reply to @conor
@conor @gregkh Thanks for the clarification. I'm pretty early stages on learning how whole RISC-V ecosystem works :-)
0
0
1
About social.kernel.org

Terms of service

  1. Please do not use this service in violation of the Linux Kernel Code of Conduct. Doing so will result in your account suspension with the referral of the matter to the CoC committee.

  2. "Repeating"/"boosting" someone else's status on this platform will be treated as endorsement and will fall under rule #1.

  3. You are encouraged to use this platform to promote your work on the Linux Kernel, but there is no restriction on permitted topics (with the exception of anything covered by #1 above).

  4. There is no requirement to post in English, but it should be considered the primary language of communication on this platform.

Privacy notice

The admins of this service have access to all posted statuses. They aren't looking, but if it's something they shouldn't know about, then you should not post it on this platform.

Please see the Linux Foundation Privacy Policy, which applies to this platform as well.

Getting your own account

If you would like an account on this instance, please check that the following applies to you:

  • You are listed in MAINTAINERS or CREDITS

  • OR: You have a kernel.org account or email address

  • OR: You have a long and established history of involvement with the Linux Kernel

If the above is true and you agree with the Terms of Service and Privacy Notice listed above, please use these instructions to request an account: